{"id":71880,"date":"2026-07-29T09:30:00","date_gmt":"2026-07-29T07:30:00","guid":{"rendered":"https:\/\/symmy.com\/erpedie\/jak-pripojit-claude-k-pohode-2\/"},"modified":"2026-08-18T07:42:31","modified_gmt":"2026-08-18T05:42:31","slug":"jak-pripojit-claude-k-pohode","status":"publish","type":"erpedie","link":"https:\/\/symmy.com\/en\/erpedie\/jak-pripojit-claude-k-pohode\/","title":{"rendered":"How to connect Claude to POHODA: a step-by-step guide"},"content":{"rendered":"<p>If your company already uses Claude \u2014 for writing, analysis or coding \u2014 it is a shame that it cannot see into your accounts. Yet of all AI assistants, Claude has the shortest path to company data: the MCP protocol it uses to connect to POHODA was created by its own makers. In this guide we cover what you need, how to get connected in a few minutes and what to check on the security side.<\/p>\n<h2>Why Claude specifically?<\/h2>\n<div class=\"ep-answer\">\n<p>Claude supports MCP natively and has done so longer than any other AI assistant \u2014 the standard was created by Anthropic, the maker of Claude. Just give it the address of an MCP server running over POHODA in its settings and you can ask questions about your data in plain language. No programming, no exports.<\/p>\n<\/div>\n<p>Anthropic introduced MCP (Model Context Protocol) at the end of 2024 as an open standard for connecting AI to business systems; today it is supported by ChatGPT and Microsoft too. For Claude, though, MCP is home turf: you set the connector up in both the web and desktop apps, and a connection added once works across all your conversations. What an MCP server actually does is <a href=\"\/en\/erpedie\/co-je-mcp-server\/\">covered in a separate article<\/a>.<\/p>\n<h2>What you will need<\/h2>\n<ul class=\"ep-cards\">\n<li><strong>A paid Claude account<\/strong>Custom connectors are available on paid plans; for companies we recommend Team or Enterprise for central management of both connectors and data.<\/li>\n<li><strong>POHODA with data access<\/strong>Direct data reading requires the SQL or E1 edition (they run on SQL Server). With the basic edition, the route goes through <a href=\"\/en\/erpedie\/pohoda-mserver\/\">mServer<\/a>.<\/li>\n<li><strong>An MCP server for POHODA<\/strong>The middleman between Claude and your data \u2014 hosted, such as <a href=\"\/en\/erp-mcp\/pohoda\/\">POHODA MCP<\/a>, or your own open-source one.<\/li>\n<li><strong>A few minutes<\/strong>The connection itself is one entry in the settings and one sign-in.<\/li>\n<\/ul>\n<h2>Connecting in three steps<\/h2>\n<ol class=\"ep-steps\">\n<li><strong>Open Settings \u2192 Connectors<\/strong> in the Claude web or desktop app and choose to add a custom connector.<\/li>\n<li><strong>Paste the MCP server address<\/strong> (e.g. <code>https:\/\/mcp.symmy.ai\/pohoda<\/code>) and sign in \u2014 it uses OAuth, so no POHODA passwords and no shared keys.<\/li>\n<li><strong>Ask in plain language.<\/strong> On first use, Claude asks you to approve each tool; from then on it answers with figures straight from your POHODA.<\/li>\n<\/ol>\n<h2>What to do with Claude over POHODA<\/h2>\n<ul class=\"ep-cards\">\n<li><strong>Receivables<\/strong>\u201cWhich invoices are overdue and by how many days?\u201d \u201cDraft a payment reminder for our three biggest debtors.\u201d<\/li>\n<li><strong>Reports and briefings<\/strong>\u201cCompare this year&#8217;s and last year&#8217;s revenue by month.\u201d \u201cPrepare the figures for last month&#8217;s VAT return.\u201d<\/li>\n<li><strong>Stock<\/strong>\u201cWhat is below minimum stock level?\u201d \u201cWhich items haven&#8217;t sold even once in six months?\u201d<\/li>\n<li><strong>Sales<\/strong>\u201cSummarise the history of company ABC before tomorrow&#8217;s meeting \u2014 orders, payments, open quotes.\u201d<\/li>\n<\/ul>\n<p>The power is in the combination: Claude doesn&#8217;t just pull the data out, it works with it on the spot \u2014 drafting a reminder, building a table, flagging an anomaly.<\/p>\n<h2>Claude vs. ChatGPT over an ERP<\/h2>\n<div class=\"ep-note\">\n<h2>Both speak MCP \u2014 you don&#8217;t have to choose<\/h2>\n<p>One MCP server serves both assistants; the only differences are where you set the connector up and the plan terms. Claude&#8217;s MCP support is the longest-standing and the deepest; ChatGPT added connectors later. If your company runs on ChatGPT, we have <a href=\"\/en\/erpedie\/jak-pripojit-chatgpt-k-pohode\/\">the same guide for it<\/a> \u2014 and feel free to connect both to the same server.<\/p>\n<\/div>\n<h2>Security: four questions before you go live<\/h2>\n<p>The same rules apply as for any AI connection to company data. Insist on a <strong>read-only default mode<\/strong>, the ability to <strong>restrict the agendas<\/strong> the assistant can see (never payroll), an <strong>audit log<\/strong> of every query, and a clear answer to the question of <strong>where the data is processed<\/strong> \u2014 the MCP server should run in the EU and not store your data. On business plans, Anthropic does not use conversation content to train its models.<\/p>\n<h2>Frequently asked questions<\/h2>\n<details class=\"ep-faq\">\n<summary>Do I need a paid Claude plan?<\/summary>\n<p>Yes, custom connectors are part of the paid plans. For individuals, Pro is enough; for companies we recommend Team or Enterprise, where an administrator sets the connector up once for the whole team.<\/p>\n<\/details>\n<details class=\"ep-faq\">\n<summary>Does the connection work in the mobile app too?<\/summary>\n<p>The connector is tied to your account \u2014 you add it in the web or desktop app and can then query your data anywhere you are signed in.<\/p>\n<\/details>\n<details class=\"ep-faq\">\n<summary>Will Anthropic see my accounts?<\/summary>\n<p>Claude only receives the data the MCP server returns for a specific query \u2014 never the whole database. On business plans, conversation content is not used to train models. The MCP server itself should run in the EU and not store data; ask your provider about that.<\/p>\n<\/details>\n<details class=\"ep-faq\">\n<summary>Can I connect an ERP other than POHODA the same way?<\/summary>\n<p>Yes \u2014 the procedure is identical, only the server address changes. We offer MCP servers for <a href=\"\/en\/erp-mcp\/\">Money S4\/S5, Helios iNuvio, KARAT, ABRA and other systems<\/a>.<\/p>\n<\/details>\n","protected":false},"excerpt":{"rendered":"<p>Claude speaks the MCP protocol natively \u2014 you can connect it to POHODA in minutes. What you need, a three-step walkthrough, and the security questions to ask.<\/p>\n","protected":false},"author":0,"featured_media":67161,"template":"","class_list":["post-71880","erpedie","type-erpedie","status-publish","has-post-thumbnail","hentry"],"acf":[],"_links":{"self":[{"href":"https:\/\/symmy.com\/en\/wp-json\/wp\/v2\/erpedie\/71880","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/symmy.com\/en\/wp-json\/wp\/v2\/erpedie"}],"about":[{"href":"https:\/\/symmy.com\/en\/wp-json\/wp\/v2\/types\/erpedie"}],"version-history":[{"count":1,"href":"https:\/\/symmy.com\/en\/wp-json\/wp\/v2\/erpedie\/71880\/revisions"}],"predecessor-version":[{"id":71951,"href":"https:\/\/symmy.com\/en\/wp-json\/wp\/v2\/erpedie\/71880\/revisions\/71951"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/symmy.com\/en\/wp-json\/wp\/v2\/media\/67161"}],"wp:attachment":[{"href":"https:\/\/symmy.com\/en\/wp-json\/wp\/v2\/media?parent=71880"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}